INSERTION OF A FUNCTION BELONGING TO A CERTAIN SUBCLASS OF R X

Transcription

1 Bulletin of the Iranian Mathematical Society, Vol. 28 No. 2 (2002), pp INSERTION OF A FUNCTION BELONGING TO A CERTAIN SUBCLASS OF R X MAJID MIRMIRAN Abstract. Let X be a topological space and E(X, R) be a subset of R X with the following properties: (1) Any constant function is in E(X, R); (2) If α, β R and f, g E(X, R), then αf + βg E(X, R); (3) If (f n ) is a sequence of functions in E(X, R) and (f n ) is uniformly convergent to f, then f E(X, R); and (4) If f E(X, R) and g is a constant function, then sup{f, g} E(X, R) and inf{f, g} E(X, R). Here, necessary and sufficient conditions in terms of lower cut sets are given for the insertion of a function of E(X, R) between two comparable real-valued functions with a certain pair of a general class of properties. The class of properties is defined by being preserved when added to a function of E(X, R) and by being possessed by any constant function. 1. Introduction A property P defined relative to a real-valued function on a topological space is an E-property provided any constant function has property P and provided the sum of a function with property P and any function in E(X, R) also has property P. If P 1 and P 2 are E-properties, the following terminology is used: (i) A space X has the weak E insertion property for (P 1, P 2 ) iff for any functions g and f on X such that g f, g has property P 1 and f has property P 2, then there exists a function h in E(X, R) such that g h f. (ii) A space X has the E insertion property for (P 1, P 2 ) iff for any functions g and f on X such that This work was supported by University of Isfahan. MSC(2000): Primary 54C30, 54C35, 54C50; Secondary 26A21 Keywords: Lower cut set, Insertion of function, Completely separated. Received:14 May 2000, Revised:20 Nov c 2002 Iranian Mathematical Society. 19

2 20 Mirmiran g < f, g has property P 1 and f has property P 2, then there exists a function h in E(X, R) such that g < h < f. (iii) A space X has the strong E insertion property for (P 1, P 2 ) iff for any functions g and f on X such that g f, g has property P 1 and f has property P 2, then there exists a function h in E(X, R) such that g h f and such that if g(x) < f(x) for any x in X, then g(x) < h(x) < f(x). In this paper for a space X with the weak E insertion property for (P 1, P 2 ), we give a necessary and sufficient conditions in terms of lower cut sets for the space to have the E insertion property for (P 1, P 2 ). Also for a space with the weak E insertion property, we present a necessary and sufficient conditions for the space to have the strong E insertion property. 1. Examples of E-property and E-insertion In the case that E(X, R) is the set continuous real-valued functions on X i.e. E(X, R) = C(X, R), then lower semicontinuous (lsc), upper semicontinuous (usc), continuity are examples of E properties. The following examples of C insertion are known: (a) A spacex has the weak C insertion property for (usc, lsc) iff X is normal. (b) A space X has the C insertion property for (usc, lsc) iff X is normal and countably paracompact. (c) A space X has the strong C insertion property for (usc, lsc) iff X is perfectly normal. Result (a) is due independently to Tong [10] and to Katětov[4]. Example (b) was proved by Katětov[4] and by Dowker [3]. Result (c) is a consequence of Theorem 3.1, Example 1.2, and Proposition 2.3 of Michael [8]. Also, we can choose E(X, R) = B 1 (X, R); the set Baire-one realvalued functions on X, or E(X, R) = A(R, R); the set approximately continuous real-valued functions on R. We now give the appropriate definitions and terminologies as follows:

3 Insertion of a function in R X 21 DEFINITION 1.1. A real-valued function f on X is called upper (resp. lower) semibaire-one, if for any real number t, the set {x X : f(x) < t} (resp. {x X : f(x) > t}) is a F σ subset of X. We denote upper semibaire-one by usb 1 and lower semibaire-one by lsb 1. In the case that E(X, R) = B 1 (X, R), then lsb 1, usb 1 and B 1 are examples of E-properties. If a space has the strong E insertion property for (P 1, P 2 ), then it has the weak E insertion and the E insertion property for (P 1, P 2 ). In order to see when the E insertion property implies the weak E ins ertion property, a technique used by Dieudonně[2] is employed to prove the following result. THEOREM 1.2. Let P 1 and P 2 be E properties and assume that X satisfies the E insertion property for (P 1, P 2 ). If (a ) the transformation f f/(1 + f ) preserves P 1 and P 2, and transformation f f/(1 f ) preserves E for any f E(X, ( 1, 1)), and (b) if inf(f, h) has property P 2 and sup(g, h) has property P 1 whenever f has property P 2, g has property P 1 and h is any function in E(X, R), then X satisfies the weak E insertion propery for (P 1, P 2 ). Proof. Let g and f be functions on X such that g f, g has property P 1 and f has property P 2. If G = g/(1+ g ) and F = f/(1+ f ), then G has property P 1 and F has property P 2 and 1 < G F < 1. If G 0 = G 1 and F 0 = F +1, then by hypothesis there exists a function h 0 E(X, R) such that G 0 < h 0 < F 0. Let f 1 = inf(f + 1/2, h 0 + 1/2), g 1 = sup(g 1/2, h 0 1/2). Then g 1 < f 1 and by hypothesis, g 1 has property P 1 and f 1 has property P 2. Inductively, let f n = inf(f + 1/2 n, h n 1 + 1/2 n ), g n = sup(g 1/2 n, h n 1 1/2 n ), where h n 1 E(X, R) and g n 1 < h n 1 < f n 1. Again by hypothesis there exists a function h n E(X, R) such that g n < h n < f n. Since h n < f n h n 1 + 1/2 n and h n > g n h n 1 1/2 n, then 1/2 n h n h n 1 1/2 n. Since h n h n 1 1/2 n, the sequence (h n ) converges uniformly to H E(X, R) by the Cauchy condition and

4 22 Mirmiran the properties of E(X, R). Since G 1/2 n g n < h n < f n F + 1/2 n and since (h n ) converges to H, it follows that G H F. Since the function t t/(1 t ) on ( 1, 1) is increasing, thus if h = H/(1 H ) then g h f and h E(X, R). Thus X satisfies the weak E- insertion property for (P 1, P 2 ). 2. E-insertion If f is a real-valued function defined on a space X and if {x : f(x) < t} A(f, t) {x : f(x) t}, for a real number t, then A(f, t) is a lower cut set in the domain of f at the level t. This definition is due to Brooks [1], where the terminology lower indefinite cut set is used. The main result of this section uses lower cut sets and gives a necessary and sufficient condition for a space that satisfies the weak E-insertion property to satisfy the E insertion property. THEOREM 2.1. Let P 1 and P 2 be E-property and X be a space that satisfies the weak E-insertion property for (P 1, P 2 ). Also assume that g and f are functions on X such that g < f, g has property P 1 and f has property P 2. The space X has the E-insertion property for (P 1, P 2 ) iff there exist lower cut sets A(f g, 3 n+1 ) and there exists a decreasing sequence (D n ) of subsets of X with empty intersection and such that for each n, X \ D n and A(f g, 3 n+1 ) are completely separated by functions in E(X, R). Proof. Assume that X has the weak E-insertion property for (P 1, P 2 ). Let g and f be functions such that g < f, g has property P 1 and f has property P 2. By hypothesis there exist lower cut sets A(f g, 3 n+1 ) and there exists a sequence (D n ) such that D n = and such that for each n, X \ D n and A(f g, 3 n+1 ) are completely separated by functions in E(X, R). Let k n be a function in E(X, [0, 1]) such that

5 Insertion of a function in R X 23 k n = 0 on A(f g, 3 n+1 ) and k n = 1 on X \ D n. Let a function k on X be defined by k(x) = 1/2 3 n k n (x). By the Cauchy condition and the properties of E(X, R), the function k is in E(X, R). Since D n = and since k n = 1 on X \ D n, it follows that 0 < k. Also 2k < f g: In order to see this, observe first that if x is in A(f g, 3 n+1 ), then k(x) 1/4(3 n ). If x is any point in X, then x / A(f g, 1) or for some n, x A(f g, 3 n+1 ) A(f g, 3 n ); in the former case 2k(x) < 1, and in the latter 2k(x) 1/2(3 n ) < f(x) g(x). Thus if f 1 = f k and if g 1 = g+k, then g < g 1 < f 1 < f. Since P 1 and P 2 are E-properties, then g 1 has property P 1 and f 1 has property P 2. Since X has the weak E-insertion property for (P 1, P 2 ), then there exists a function h E(X, R) such that g 1 h f 1. Thus g < h < f, it follows that X satisfies the E-insertion property for (P 1, P 2 ). (The technique of this proof is by Katětov[4]). Conversely, let g and f be functions on X such that g has property P 1, f has property P 2 and g < f. By hypothesis, there exists a function h E(X, R) such that g < h < f. We follow an idea contained in Lane [6]. Since the constant function 0 has property P 1, since f h has property P 2, and since X has the E-insertion property for (P 1, P 2 ), then there exists a function k E(X, R) such that 0 < k < f h. Let A(f g, 3 n+1 ) be any lower cut set for f g and let D n = {x X : k(x) < 3 n+2 }. Since k > 0 it follows that D n =. Since A(f g, 3 n+1 ) {x X : (f g)(x) 3 n+1 } {x X : k(x) 3 n+1 } and since {x X : k(x) 3 n+1 } and {x X : k(x) 3 n+2 } = X \ D n are completely separated by sup{3 n+1, inf{k, 3 n+2 }} E(X, R), it follows that for each n, A(f g, 3 n+1 ) and X \ D n are completely separated by functions in E(X, R). 3. Strong E-insertion The main result of this section uses lower cut sets and gives a necessary and sufficient condition for a space that satisfies the weak E insertion property to satisfy the strong E insertion property.

6 24 Mirmiran THEOREM 3.1. Let P 1 and P 2 be E-property and X be a space that satisfies the weak E-insertion property for (P 1, P 2 ). Also assume that g and f are functions on X such that g f, g has property P 1 and f has property P 2. The space X has the strong E-insertion property for (P 1, P 2 ) iff there exists a sequence (A(f g, 2 n )) of lower cut sets for f g and there exists a sequence (F n ) of subsets of X such that (i) {x X : (f g)(x) > 0} = F n, and (ii) for each n, the sets A(f g, 2 n ) and F n are completely separated by functions in E(X, R). Proof. Suppose that there is a sequence (A(f g, 2 n )) of lower cut sets for f g and suppose that there is a sequence (F n ) of subsets of X such that {x X : (f g)(x) > 0} = F n and such that for each n, there exists a function k n E(X, [0, 2 n ]) with k n = 2 n on F n and k n = 0 on A(f g, 2 n ). The function k from X into [0, 1/4] which is defined by k(x) = 1/4 k n (x) is in E(X, R) by the Cauchy condition and the properties of E(X, R), (1) k 1 (0) = {x X : (f g)(x) = 0} and (2) if (f g)(x) > 0 then k(x) < (f g)(x) : In order to verify (1), observe that if (f g)(x) = 0, then x A(f g, 2 n ) for each n and hence k n (x) = 0 for each n. Thus k(x) = 0. Conversely, if (f g)(x) > 0, then there exists an n such that x F n and hence k n (x) = 2 n. Thus k(x) 0 and this verifies (1). Next, in order to establish (2), note that {x X : (f g)(x) = 0} = A(f g, 2 n ) and that (A(f g, 2 n )) is a decreasing sequence. Thus if (f g)(x) > 0 then either x A(f g, 1/2) or there exists a smallest n such that x A(f g, 2 n ) and x A(f g, 2 j ) for j = 1,..., n 1. In the former case, k(x) = 1/4 k n (x) 1/4 2 n < 1/2 (f g)(x),

7 Insertion of a function in R X 25 and in the latter, k(x) = 1/4 k j (x) 1/4 2 j < 2 n (f g)(x). j=n j=n Thus 0 k f g and if (f g)(x) > 0 then (f g)(x) > k(x) > 0. Let g 1 = g + (1/4)k and f 1 = f (1/4)k. Then g g 1 f 1 f and if g(x) < f(x) then g(x) < g 1 (x) < f 1 (x) < f(x). Since P 1 and P 2 are E-properties, then g 1 has property P 1 and f 1 has property P 2. Since by hypothesis X has the weak E-insertion property for (P 1, P 2 ), then there exists a function h E(X, R) such that g 1 h f 1. Thus g h f and if g(x) < f(x) then g(x) < h(x) < f(x). Therefore X has the strong E-insertion property for (P 1, P 2 ). (The technique of this proof is by Lane [6].) Conversely, assume that X satisfies the strong E-insertion for (P 1, P 2 ). Let g and f be functions on X satisfying P 1 and P 2 respectively such that g f. Thus there exists h E(X, R) such that g h f and such that if g(x) < f(x) for any x in X, then g(x) < h(x) < f(x). We follow an idea contained in Powderly [9]. Now consider the functions 0 and f h.0 satisfies property P 1 and f h satisfies property P 2. Thus there exists function h 1 E(X, R) such that 0 h 1 f h and if 0 < (f h)(x) for any x in X, then 0 < h 1 (x) < (f h)(x). We next show that {x X : (f g)(x) > 0} = {x X : h 1 (x) > 0}. If x is such that (f g)(x) > 0, then g(x) < f(x). Therefore g(x) < h(x) < f(x). Thus f(x) h(x) > 0 or (f h)(x) > 0. Hence h 1 (x) > 0. On the other hand, if h 1 (x) > 0, then since (f h) h 1 and f g f h, therefore (f g)(x) > 0. For each n, let A(f g, 2 n ) = {x X : (f g)(x) 2 n }, F n = {x X : h 1 (x) 2 n+1 } and k n = sup{inf{h 1, 2 n+1 }, 2 n } 2 n. Since {x X : (f g)(x) > 0} = {x X : h 1 (x) > 0}, it follows that {x X : (f g)(x) > 0} = F n.

8 26 Mirmiran We next show that k n is in E(X, [0, 2 n ]) which completely separates F n and A(f g, 2 n ). From its definition and by the properties of E(X, R), it is clear that k n is in E(X, [0, 2 n ]). Let x F n. Then, from the definition of k n, k n (x) = 2 n. If x A(f g, 2 n ), then since h 1 f h f g, h 1 (x) 2 n. Thus k n (x) = 0, according to the definition of k n. Hence k n completely separates F n and A(f g, 2 n ). THEOREM 3.2. Let P 1 and P 2 be E-properties and assume that the space X satisfied the weak E-insertion property for (P 1, P 2 ). The space X satisfies the strong E-insertion property for (P 1, P 2 ) iff X satisfies the strong E-insertion property for (P 1, E) and for (E, P 2 ). Proof. Assume that X satisfies the strong E-insertion property for (P 1, E) and for (E, P 2 ). If g and f are functions on X such that g f, g satisfies property P 1, and f satisfies property P 2, then since X satisfies the weak E-insertion property for (P 1, P 2 ) there is a function k E(X, R) such that g k f. Also, by hypothesis there exist functions h 1 and h 2 in E(X, R) such that g h 1 k and if g(x) < k(x) then g(x) < h 1 (x) < k(x) and such that k h 2 f and if k(x) < f(x) then k(x) < h 2 (x) < f(x). If a function h is defined by h(x) = (h 2 (x) + h 1 (x))/2, then h is in E(X, R), g h f, and if g(x) < f(x) then g(x) < h(x) < f(x). Hence X satisfies the strong E-insertion property for (P 1, P 2 ). The converse is obvious since any function in E(X, R) must satisfy both properties P 1 and P 2. (The technique of this proof is by Lane [7].) Remark. In conclusion, let us mention that, in the case that E(X, R) = C(X, R), Theorems 1.2, 2.1, 3.1 of Lane [6] and Proposition 2.1 of Lane [7] are respectively consequences of our Theorems 1.1, 2.1, 3.1 and 3.2. References [1] F. Brooks, Indefinite cut sets for real functions, Amer. Math. Monthly, 78(1971), [2] J.Dieudonně, Une generalisation des espaces compacts, Journal de Math. Pures et Appliquěs, 23(1944), [3] C. H. Dowker, On countably paracompact spaces, Canad. J. Math., 3(1951),

9 Insertion of a function in R X 27 [4] M. Katětov, On real-valued functions in topological spaces, Fund. Math., 38(1951), [5] M. Katětov, Correction to, On real-valued functions in topological spaces, Fund. Math., 40(1953), [6] E. Lane, Insertion of a continuous function, Pacific J. Math., 66(1976), [7] E. Lane, PM-normality and the insertion of a continuous function, Pacific J. Math., 82(1979), [8] E. Michael, Continuous selection I, Ann. of Math., 63(1956), [9] M. Powderly, On insertion of a continuous function, Proc. Amer. Math. Soc., 81(1981), [10] H. Tong, Some characterizations of normal and perfectly normal spaces, Duke Math. J., 19 (1952), Department of Mathematics University of Isfahan Isfahan 81744, Iran. mirmir@sci.ui.ac.ir